Abstract: Aimed at the weak self-learning ability of traditional expert system, the expression method for fuzzy subordination and the mode of fuzzy back-propagation artificial neural network for 26 kinds of common grape diseases was studied, so as to realize a web based intelligent grape disease diagnosis system, which was implemented by JAVA and MATLAB. The system can be popularized easily as it is designed to be run online, and experimental results show that the system can diagnose grape diseases with the accuracy of 90.9%. The analyses of the diagnosis results of typical examples indicate that this system has stable reliability, can simulate the expert diagnosis process adequately, and can improve the diagnosis efficiency greatly.